"AMI's 0.8m Gate Array family is simply the best 0.8m on the market . . . one of the highest performance, yet lowest cost array products available today . . ." Designed for 3V, 5V, or 3V/5V mixed supplies 210 ps gate delays (fanout = 2) 5,000 to 663,000 available gate densities Complete package lineup quad flatpack (QFP), LCC, DIP, grid array . . . Usable megafunctions Families of single port RAMs, microprocessors, controllers, datapath functions . . . "Made in America" engineering, manufacturing, and support No `overseas' delays to your important questions; we're right here, ready to help. Table of Contents Features....................................................................................... 1 AMI8Gx Gate Array Family ............................................................ 1 Architectural Overview .................................................................. 2 Product Applications ..................................................................... 3 ASIC Design Tools and Methodology ............................................. 3 The Design Library ........................................................................ 5 DC Specifications ........................................................................ 6 Library Cell Selection Guide ................................................... 8 - 21 Delay Derating Information ............................................................ 22 Packaging .................................................................................... 23 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 AMI's "AMI8Gx" series of 0.8m gate arrays exploits a proprietary power grid and track routing architecture on a compact, channelless, sea-of-gates design to provide one of the highest performance, cost effective array products available today. * JTAG Boundary Scan macro support * Cost driven architecture: - Offers both 2 and 3 level metal interconnect to provide the lowest user cost for the number of gates and pads required. - Provides 6 extra power pads per corner to preserve more I/O cells for signal use. - Contains extra I/O cells to provide extra drive without wasting bond pads. Features * 3V, 5V, and combined 3V/5V operation: Each individual pad cell can be driven independently by a 3V or 5V supply. 3V to 5V and 5V to 3V level shift is available in all I/O cells. Core can be either 3V for low power or 5V for high speed. * Extensive library for quick design: - Complete primary cell and I/O library. - Synchronous single port RAM compilers with over 2000 compiled RAM sizes from 32x1 to 1Kx32 bits. - MG65C02, MG29C01, MG29C10, MG80C85, MG82Cxx, MGMC51 megacells. - Various datapath logic synthesizers (FIFOs, multipliers, adders, barrel shifters). * Operating Temp equals -55 to 125 oC: Few competing products allow this range. * Excellent performance: - 260 MHz maximum toggle rate on clocked flip-flops (TJ = 135 oC). - 200 ps delay (FO=2) for a 2-input NAND gate. Clock tree generation: 400 ps clock skew (fan out = 3500 at 80 MHz). * 1 to 16 mA drive per single I/O cell: Selectable I/O drive with controllable slew rate. Extra I/O cells allow combined I/O drive up to 96mA without reducing pad count. * Wide range of packaging: Full QFP and LCC line, DIPs and PGAs, individual die, (ball grid array package under study). Burn-in capability as needed. * Automatic Test Program Generation: Includes scan macros (NETSCAN) for high fault coverage. * Power equals 3.2 W/MHz/cell * Full operating voltage range from 2.7V to 5.5V * ESD protection > 2kV; latchup > 100mA AMI8Gx Gate Array Family Part Number Raw Gates Usable Gates1 Triple Metal Double Metal Available I/O Cells Available Bond Pads2 AMI8G663 663,350 464,350 331,700 732 528 AMI8G392 392,160 274,500 196,100 564 424 AMI8G247 246,790 172,750 123,400 448 336 AMI8G201 200,740 140,500 100,400 404 304 AMI8G142 141,590 99,100 70,800 340 256 AMI8G93 92,740 64,900 46,400 276 208 AMI8G65 65,420 45,800 32,700 232 176 AMI8G55 54,700 38,300 27,400 212 160 AMI8G44 44,460 31,100 22,200 192 144 AMI8G34 33,860 23,700 16,900 168 128 AMI8G21 20,800 14,550 10,400 132 100 AMI8G15 14,960 10,470 7,480 112 84 AMI8G9 8,900 6,230 4,450 88 68 AMI8G5 5,250 3,670 2,630 68 52 Note1:Exactusablegatecountwillvarydependingondesigninterconnectandmacroselection. Note2:24optionalfixedpowerpins(6ineachcorner)arenotincludedinthisnumber. 1 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 FIGURE 1: GATE ARRAY ARCHITECTURE Selectable 2 or 3 metal interconnect [note 1] Large I/O driver with prebuffer [note 5] Core cell sites with 4 transistors [note 2] 2 P-channel 2 N-channel Extra corner power pads [note 4] Separately configured power busses for I/O and core [note 3] Architectural Overview Some important elements of the AMI8Gx gate array family are: * [Note 1] Drawn gate length of 0.8 micron; 2 or 3 level metal interconnect selectable. * [Note 4] Twenty-four (six per corner) fixed power pads (not included in Bond Pad count in Table 1) available on each array for customer use. * Each I/O cell can be configured as 5V VDD, 3V VDD, VSS, or signal I/O. * [Note 2] Two p-channel and two n-channel transistors per site (or cell). Sites are arrayed in a sea-of-gates structure that can allow interconnect routing over active sites. Also, p-channel transistors are sized larger than the stronger n-channel transistors in each cell to provide better matched rise times and fall times. * [Note 5] Each I/O cell has selectable drive from 1mA to 16mA. All I/O cell logic can be built in the I/O cell prebuffer. Level shifting (3V to 5V or 5V to 3V) may require the use of a few core gates. * [Note 3] Four separate power busses for I/O cells to allow separate supplies for output buffers, input buffers, and mixed VDD levels all on an individual I/O cell basis. Two separate power busses for core logic (not shown). 2 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 Product Applications known layouts are used. Once actual layout is completed by AMI, a post-layout interconnect capacitance table will be supplied for final validation of device timing. The family's extended temperature and voltage operation range make it well suited for telecom, industrial, and military applications. The low cost structure also makes it ideal in computer and office automation ASIC requirements. Figure 2 shows a typical design flow for a new design. FPGA OR PAL CONVERSION: AMI can convert netlists from most FPGA and PAL devices to a more cost and performance effective AMI8Gx gate array design for volume production. 2ND SOURCE EXISTING PRODUCTS: Netlist conversion capabilities from AMI allow a competitive alternate supply with AMI8Gx for current high volume designs. NEW DESIGN CAPTURE: AMI8Gx design is supported by many popular third party software platforms, as well as AMI's Enhanced Design Utilities (EDU) environment. PROCESS UPGRADE: Designs done in AMI's 1.25m and 1.0m gate array families can easily be upgraded to the AMI8Gx family. The AMI ASIC Standard Library provides a common netlist design base. ASIC Design Tools and Methodology AMI8Gx and other AMI ASIC families are supported on popular third party products: * * * * * * * Cadence Mentor Graphics Synopsys Viewlogic Valid Verilog simulation IKOS simulation accelerator (AMI's sign-off simulator) AMI has maintained critical proprietary software tools to ensure a tight, well coupled design to our silicon process. This methodology includes our expert-system design analysis tools, AMI's Enhanced Design Utilities (EDU), a software support methodology that covers the complete set of wafer processing possibilities, and a dedicated, experienced engineering staff that can assist at any level of the design process. AMI Design Flow AMI will supply an AMI8Gx design kit which includes a cell library containing symbols, simulation models and software for design verification, timing calculations, and netlist generation. For pre-layout timing simulations, capacitance values derived from statistical averages of 3 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 AMI Design Flow (cont.) Working with an AMI design center, the customer is responsible for capturing and verifying the design using the AMI ASIC Standard Library. He is also responsible for creating the test vectors that will eventually serve as the logical part of the manufacturing test. Software aids such as logic synthesis, megacells, automatic test program generation, netlist rule checkers, etc. can greatly speed up this process. (A fault coverage check of the test vector set is optional and can be done as an additional service.) After layout has been completed the interconnect data is extracted from the physical layout to be fed back to the sign-off simulator for final circuit verification. This post layout interconnect data can be sent to the customer for final validation on his simulator. When the post-layout simulation has been completed and approved by the customer the design is then released for mask and wafer fabrication. The test program is developed in parallel using internal automatic test program generation software. Prototypes can then be tested before they are shipped. When the design is received by the factory, the "Design Start Package" is reviewed by AMI engineers. This start package, which is completed by the customer, contains the device specification, netlist, critical timing paths, and test vectors. The design is pre-screened on the Enhanced Design Utilities (EDU) and then resimulated on IKOS, AMI's sign-off simulator. The results are compared to the customer's simulation from the third-party CAE tool. Figure 3 outlines a typical software environment when using third party tools. AMI uses EDIF to speed ports between various software products. AMI's Enhanced Design Utilities tools are intended to be used interactively at each stage of the design. EDU software is a set of design analysis tools that check both the design and test vectors for correctness and compatibility with in-house ASIC testers, and analyze the design for inefficiencies and possible flaws that could cause problems in manufacturing the device. Once the design has passed the initial screening it is then ready for placement and routing. The layout proceeds by first placing memory and megacells, assigning priority to critical paths, and designing the distribution and buffering of clocks. Next, the layout is completed with automatic place-and-route on the balance of the circuit. FIGURE 3: DESIGN ENVIRONMENT WITH THIRD PARTY SOFTWARE AMI Environment ** AMI ASIC Std. Library Logic Synthesis Memory Compiler HDL Optional Synthesis Tool ** AMI ASIC Stnd. Library Schematic Translation Physical Data Schematic Entry ** Design Database Netlist Translation Third Party Environment ** Enhanced Design Utilities supplied ** Elements in AMI Design Kit Vector Generation Estimated Delays ** Models & Symbols VHDL Timing Simulation Design Verification Place and Route Post Route Verification ATPG 4 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 Memory Compiler Library Size Memory Compiler SRAM (single-port, synchronous) min. max. 32 x 1 1K x 32 Increment 16 words, 1 bit Comments 9 ns typical access time on 1Kx16 NOTE:OtherSRAMandROMcompilersareavailableforstandardcellorembeddedarraydesignapproaches. ContactanAMIDesignCenterfordetailsabouttheseotherproductofferings. The Design Library Memory Compilers AMI provides a robust collection of building blocks for the AMI8Gx gate array family. A broad range of primary cells is complemented with memory cell compilers and useful megacells. With such broad, US-based design talent, AMI can quickly design specific cells that customers need to add an edge in customization. The AMI8Gx family offers the memory compiler shown above. Each of the thousands of possible memory blocks created by this compiler is optimized precisely to the customers' parameters rather than built from a presized leaf cell that covers a range of sizes. This yields a better size and performance match for each application. The AMI ASIC Standard Library Upon supplying the cell specification to AMI, the customer can receive an accurate simulation timing specification overnight by facsimile and a full simulation model for any AMI supported software environment within five working days. The AMI ASIC Standard Library contains a rich set of core and pad cells which allow great flexibility in building competitive devices for customer applications. The library is portable across all AMI's gate array and standard cell families. The ASIC Standard Library is listed in detail on pages 9 to 21. Datapath Synthesizers AMI8Gx also supports the complex datapath logic functions listed here. These functions are synthesized from an input set of design parameters, and can be optimized for either minimum delay, minimum area or a compromise between the two. Contact AMI for the size range and parameter set for any desired functions. Soft Datapath Library (xx by yy) Name Function MGAxxyyDv Adder MGAxxyyEv Adder-subtracter MGBxxyyAv Arithmetic/barrel shifter MGBxxBv Barrel shifter These logic synthesizers produce soft megacell schematics in the ASIC Standard Library, and a schematic symbol for incorporation and simulation with the design netlist. MGBxxCv Arithmetic shifter Megacells MGCxxAv 2-function binary comparator MGCxxBv 6-function binary comparator MGDxxAv Decrementer MGFxxyyC1 Latch-based FIFO MGIxxAv Incrementer MGIxxBv Incrementer/decrementer MGMxxyyDv Signed/unsigned multiplier MGMxxyyEv Multiplier-accumulator MGSxxyyAv Signed/unsigned subtracter The AMI8Gx gate array family supports soft megacell versions that are compatible with many popular architectures. These products are listed on the following page. Soft megacells are functionally and logically compatible with the stand alone products, but since the function is captured in a gate array, each instance of the megacell will differ slightly depending on its physical placement on the array. Soft megacells become part of the design netlist, requiring backannotation of interconnect capacitance after place-and-route for final verification. AMI supplies an actual gate level netlist and schematic of the soft megacell allowing the user to make design changes or remove unneeded features. Test vectors are provided and can be used directly or incorporated into the overall design test. All soft megacells are static designs and use AMI's ASIC Standard Library to ensure portability. 5 0.8 micron CMOS Gate Arrays Digital ASICs Products and Services August 1993 Soft Megacell Library Name Function Name Function MG1468C18 Real-time clock MG82C50A Asynchronous comm. element MG29C01 4-bit microprocessor slice MG82C54 Programmable interval timer MG29C10 Microprogram controller/sequencer MG82C55A Programmable peripheral interface MG65C02 8-bit microprocessor MG82C59A Programmable interrupt controller MG80C85 8-bit microprocessor MGMC51 8-bit microcontroller, 8051 compatible MG82C37A Programmable DMA controller MGMC51FB 8-bit microcontroller, 8051 compatible DC Specifications Operating Specifications Parameter Minimum VDD, Supply Voltage Ambient Temperature - Military - Commercial CMOS Input Specifications Maximum Units 2.7 5.5 Volts -55 125 C 0 70 C (4.5V